The feature is enabled for India users for now
What is SMS Categorizer?
- SMS Categorizer is a model that predicts the category of the SMS based on the content.
- Based on the content, the model moves messages among 3 tabs
1) Personal messages (contacts)
2) Transactional + OTP (others)
3) Spam (spam)
Why Categorization?
- Sender ID based categorization is not very effective for messages as compared to calls
- Currently, every message from a sender (say AMAZON) will go either spam (if spam score is high) or others (if no spam score). But AMAZON sends both promotions and order tracking SMSes.
- Also, to separate personal & non-personal messages (Contacts tab will have conversations).
Overview
The model is built by manually training messages (tens of thousands of messages) to our AI engine. We have trained Transactional, OTP, Spam messages (Indian English SMS content)
- The model can now identify similar messages based on training.
- The model works completely offline and does not send any messages to servers.
- Users can reclassify (feedback) a message which they find is in the wrong tab. In which case model will update itself (locally) and in future, similar messages will be classified as per the users' signals (personalized for that user).
Tabs
1) Contacts:
- Messages from your contact book
- Conversations (senders with both outgoing and incoming messages)
2) Others:
- Transactional messages
- OTP messages
- Other non-spam messages from non-contacts (if you reply to the message, the thread will move to Contacts tab)
3) Spam:
- Messages which is considered spam based on message body
- Blocked Senders
- Top spammers
Note: The same sender can have messages split between others/spam (good messages in others, bad in spam).
Report spam manually
- Long press on a message to select and hit the 3-dot menu icon on the top right to see "Report spam"
- Users' can also select multiple messages and mark as spam
- Or block the sender ID (TRVTGL) all together
Unspam a message
- From the Spam tab, Long press on a message to select and hit the 3-dot menu icon on the top right to see "Not spam", users can also select multiple messages to report.
- Users can also provide additional info by selecting the type of the message (OTP, Transaction, Others) it belongs to.
Interface
Comments
0 comments
Please sign in to leave a comment.